YASMIN: Efficient Intra-node Communication Using Generic Sockets

Nowadays, virtual machines are becoming widely used and their range of applications include a large number of scientific fields. From HPC to IaaS, communication between co-located VMs is a critical factor of efficiency. In our paper, we examine communication methods between VMs located in the same physical node, optimizing communication cost without sacrificing upper-layer API compatibility. We present YASMIN (Yet Another Shared Memory Implementation for Intra-Node), a generic socket-compliant framework for intra-node communication in the Xen hypervisor. We build on the concept of Vchan, a Xen library for intra-node communication between different VMs and we use Xen granting and signaling mechanisms to provide an efficient communication framework. The key of our design is the transport layer which runs underneath the AF_VSOCK protocol family, implemented as a dynamically inserted module. We are able to achieve 4.4x higher bandwidth rate and 65% lower latency without the need of application binary recompilation.

[1]  Nectarios Koziris,et al.  V4VSockets: low-overhead intra-node communication in Xen , 2015, CloudDP '15.

[2]  Ching-Hsien Hsu,et al.  A Xen-Based Paravirtualization System toward Efficient High Performance Computing Environments , 2010, MTPP.

[3]  Seung Ryoul Maeng,et al.  Locality-aware dynamic VM reconfiguration on MapReduce clouds , 2012, HPDC '12.

[4]  Richard E. Brown,et al.  United States Data Center Energy Usage Report , 2016 .

[5]  David Barr,et al.  Performance Enhancement of SMP Clusters with Multiple Network Interfaces Using Virtualization , 2006, ISPA Workshops.

[6]  Radu Sion,et al.  Enhancement of Xen's scheduler for MapReduce workloads , 2011, HPDC '11.

[7]  Jeremy Kepner,et al.  HPC-VMs: Virtual machines in high performance computing systems , 2012, 2012 IEEE Conference on High Performance Extreme Computing.

[8]  Christian Engelmann,et al.  Performance comparison of two virtual machine scenarios using an HPC application: a case study using molecular dynamics simulations , 2009, HPCVirt '09.

[9]  Dhabaleswar K. Panda,et al.  Virtual machine aware communication libraries for high performance computing , 2007, Proceedings of the 2007 ACM/IEEE Conference on Supercomputing (SC '07).

[10]  Dhabaleswar K. Panda,et al.  A case for high performance computing with virtual machines , 2006, ICS '06.

[11]  Jian Wang,et al.  XenLoop: a transparent high performance inter-vm network loopback , 2008, HPDC.

[12]  Raymond Namyst,et al.  Efficient Shared Memory Message Passing for Inter-VM Communications , 2008, Euro-Par Workshops.

[13]  Hai Jin,et al.  Adaptive Disk I/O Scheduling for MapReduce in Virtualized Environment , 2011, 2011 International Conference on Parallel Processing.

[14]  Xiaolan Zhang,et al.  XenSocket: A High-Throughput Interdomain Transport for Virtual Machines , 2007, Middleware.